Skip to main content
Internet Archive's 25th Anniversary Logo

Full text of "Reconstructing the Thomson Optical Depth due to Patchy Reionization with 21-cm Fluctuation Maps"

See other formats


SUBMITTED TO ApJL 

Preprint typeset using 1^1^^ style cmulatcapj v. ^jllj^A, 



RECONSTRUCTING THE THOMSON OPTICAL DEPTH DUE TO PATCHY REIONIZATION WITH 21-CM 

FLUCTUATION MAPS 

Gilbert P. Holder^Ilian T. Iliev^, Garrelt Mellema^ 

submitted to ApJL 

ABSTRACT 

Large fluctuations in the electron column density can occur during the reionization process. We 
investigate the possibility of deriving the electron density fluctuations through detailed mapping of the 
redshifted 21-cm emission from the neutral medium during reionization. We find that the electron- 
scattering optical depth and 21-cm differential brightness temperature are strongly anti-correlated, 
allowing optical depth estimates based entirely on redshifted 21-cm measurements. This should help 
to isolate the cosmic microwave background (CMB) polarization fluctuations due to reionization, 
allowing the removal of the patchy reionization polarization signal from the other polarization signals 
and a measurement of the primordial CMB quadrupole at various locations in the universe at the 
epoch of reionization. This latter application in principle allows three-dimensional mapping of the 
primordial density field at z ~ 1100 over a large fraction of the Hubble volume. 

Subject headings: radiative transfer — cosmology: theory — cosmic microwave background — intergalac- 
tic medium — large-scale structure of universe — radio lines 



1. INTRODUCTION 

Two new frontiers in astrophysics and cosmology are 
the search for small angle polarization anisotropics in the 
cosmic microwave background and the efforts to map the 
redshifted 21-cm emission from neutral hydrogen at the 
epoch of reionization. On small angular scales, a sig- 
nificant source of CMB polarization comes from Thom- 
son scattering of the primordial quadrupole anisotropy 
by compact regions of ionized gas. These ionized regions 
would appear in maps of the neutral gas as holes in a rel- 
atively uniform map. Thus we expect that where there 
are strong sources of Thomson scattering there should 
be missing 21-cm emission and where there is strong 21- 
cm emission there should be low Thomson optical depth, 
Tes, i.e. the two should be anti-correlated. In this paper 
we investigate the relation between Tes and neutral hy- 
drogen optical depth in a quantitative way that properly 
captures the relevant physics using numerical simulations 
of structure formation with radiative transfer. 

This topic is also related to the idea of Kamionkowski 
& Loeb (1997) for using polarized CMB anisotropy 
due to Thomson scattering by galaxy clusters of the 
CMB temperature quadrupole as a means to extract 
more information about the density fie ld at z ^ 1100 
than we can measure in th e CMB ( Sazo nov fc SunvaevI 
1^99; Scto & Sasaki 2000; Bamnann & Coorav '2003 
Eartsmouth 2004: S cto & Picrpaoli 2005; Bimn 20q1 
Shimon et al.l |20CI6() ^ The information content of the 
quadrupole measurements has been carefully investi- 
gated by Bunn (2006) , where it was found that higher 
redshift clusters are preferred for this, since there is less 
overlap with our me asured CMB sky. Clust er Tes are typ- 
ically around 0.005 l|Mason fc Mversll2000() . and we wiU 
show that typical fluctuations of Tcs during reionization 

^ Department of Physics, McGill University, Montreal, QC H3A 
2T8; Canada Research Chair; CIAR Scholar 

^ Canadian Institute for Theoretical Astrophysics, University of 
Toronto, 60 St. George Street, Toronto, ON M5S 3H8, Canada 

^ Stockholm Observatory, AlbaNova University Center, Stock- 
holm University, SE-106 91 Stockholm, Sweden 



can be of comparable size. The quadrupole at the time 
of reionization should not be affected by the integrated 
Sachs- Wolfe effect and thus provides a remarkably clean 
measure of very large scale structure at z ~ 1100. 

Reionization occurs in a patchy way: strongly- 
clustered sources create large ionized regions embedded 
in a mainly neutral medium, which generates CMB po- 
larization anisotropics on small scales. Scattering of the 
primordial CMB temperature quadrupole leads to lin- 
early polarized signals from ionized regions, where the 
size of the ionized bubble sets the scale of the anisotropy. 
On large scales, the relevant quantity is just the mean 
ionization as a function of redshift, since the small scale 
structure is averaged out. This is the signal that WMAP 
detected as the hallmark of reionization. 

The amplitude of the expected signal is expected to 
be on the order of the fluctuations in the electron scat- 
tering optical depth times the amplitude of the rele- 
vant primordial quadrupole. The latter should be on 
the order of 15 /iK but the former is not well con- 
strained at present. Calculations done assuming that 
the ionization fluctuations can be treated as a Gaussian 
random field find very small fluctuations, since modes 
along the line of sight (LOS) tend to ave rage down l|Hiil 
1 20001). However, rece nt simulations t I hev et all l2006lf 
iMellema et all l2006bfl show surprisingly large fluctua- 
tions in the optical depth, on the order of 0.01 coming 
just from the epoch of reionization, as we will show be- 
low. This optical depth is larger than that found in even 
the largest galaxy clusters in the local universe. The all- 
or-nothing nature of the ionized regions effectively breaks 
the Limber approximation, often leading to very little 
cancellation along the LOS. Recent work has suggested 
that patchy reionization scenari os can result in interest- 
ing levels of CMB polarization l)Mortonson fc Hu 2006). 

In principle, one can use a H I map to synthesize an 
effective Tes map and compare with a CMB polarization 
map to infer the CMB quadrupole at the location of the 
scattering. Naively, the fluctuation maps (after subtrac- 
tion of the spatial means) should differ only by an overall 



2 



amplitude scaling, provided the primordial quadrupole is 
slowly varying over the epoch of reionization. 

We assume a flat ACDM cosmology with parame- 
ters (nrr.,nx rih^Kaf^.n) = (0.27,0.73,0.044,0.7,0.9,1) 
l)SDergel200.1ft . where flm, JIa, and fit, are the total mat- 
ter, vacuum, and baryonic densities in units of the critical 
density, erg is the present rms linear density fluctuation 
on the scale of 8/i~^Mpc, and n is the index of the pri- 
mordial power spectrum of density fluctuations. 

2. SIMULATIONS 

Our bas i c m ethodology was described in detail in 
llliev et al] l)2006b..) . We start by performing a high reso- 
lution 100 h~^Mpc N-body simulation with a spatial grid 
of 3248^ cells, 1624^ = 4.3 billi on particles, and using the 
particle-mesh code PMFAST l|Merz et al.ll2005|) . This 
yields detailed halo catalogues, and density and veloc- 
ity fields at up to 100 roughly equally-spaced times. All 
halos identified in our simulation volume are assumed 
to be sources of ionizing radiation with ionizing pho- 
ton emissivity given by a constant mass-to-light ratio. 
We follow the time-dependent propagation of the ioniza- 
tion fronts produced by sources in the simulation volume 
using a detailed radiative transf er and non-equilibrium 
chemistry code called C^-Ray ()Mellema et al.l 1200631) . 
which has been extensively tested agai nst available ana- 
lytical solutions ijMellema et alJl2006af) and a number of 
other cosniological radiative transfer codes (jlliev et al.l 
I2006aj) . The transport of ionizing radiation is done on a 
coarsened grid of 203^ or 406^ cells, in order to make 
the problem tractable. These simulatio ns allow for a 
number of detailed predictions of 21-cm (iMellema et al.l 
l2006bj and kSZ signals (Ilicv et al. 2006c, ; Ihev et al. in 
prep .). In this work we us e data from simulation f2000 
fsee IMellema et aLll2n06bj) . though our conclusions are 
generic. 

3. FLUCTUATIONS FROM PATCHY REIONIZATION 

3.1. 21-cm Emission Fluctuations 

For our assumed cosmology (where dark energy and 
curvature are dynamically unimportant at the epoch of 
reionization) and ignoring the peculiar motions (i.e. as- 
suming that the line profile is determined solely by the 
Hubble expansion) the 21-cm optical depth of a hydrogen 
cloud can be written as 

Shc^Aio nm{z) 

t{v)- 



^ 8.5 mKxHi (1 + 5p) (1 + zf/"^. 



(2) 



■i2'KkBTsvl H{z) 
8.5mK 



(1) 



where Aiq — 2.85 x 10^^^ is the Einstein A-coefficient 
for spin-flip transition, T5 is the 21-cm transition spin 
temperature, vq is the rest-frame line frequency, H(z) — 
+ z)^^^ is the Hubble constant at redshift z, 
and nHi(z) is the hydrogen number density, xm is the 
mass- weighted neutral fraction of hydro gen and Sp is the 
local overdensity (e.g. llliev et al .' 2002"). For simplicity, 
we assume that the spin temperature is much greater 
than the CMB temperature, which should be a good ap- 
proximation after the earliest phases of reionization, and 
that T <C 1, in which case the 21-cm differential bright- 
ness temperature is given by: 

AT,(z.) = ~ ^""'^ [1 - e-^^--)] 



We can express the neutral fraction in terms of the 
mean ionized fraction and fluctuations around the mean 



xhi = 1 - x„ 



Sx, 



(3) 



where Xm is the mean mass- weighted ionized fraction and 
Sx is the fluctuation in the ionized hydrogen fraction. 
Combining equations Q and Q yields 



Nl/2 



x[l~X,n + {l-Xm)Sp-Sx{l+Sp)]. (4) 

If we subtract off the spatial mean, as most observing 
schemes necessarily would, then the fluctuations in the 
21-cm emission are given by 

S[ATb{iy)] w 8.5 mK{l + zfl'^ 

X [(l-i„)(5p-(5a;(l + (5p)]. 

3.2. Thomson optical depth fluctuations 
The optical depth to Thomson scattering is 



(5) 



ne{z) <tt cdt, 



(6) 



where gt is the Thomson cross-section. There is some 
subtlety here in relating this to the hydrogen distribu- 
tion, since some of the free electrons come from ionized 
helium. For simplicity we assume that fluctuations in 
the helium ionization trace the hydrogen ionization. Ex- 
treme variations in the ionization state of helium could 
change the optical depth fractionally at the 10% level. 
Using A^Hc to denote the number of electrons from he- 
lium ionizations per hydrogen ionization 



(l + AfHc)nH(0)(l + z)VT 



H{z) 



X [xm{l + 5p) + 5x{l + 5p)\dz. 
Subtracting the spatially uniform part we find 
,5res = 0.0031(1 + ArHo) 

X / dz[xm5p + 5x{l + 5p)]{l + zf/'^. 



(7) 



(8) 



4. COMBINING DATA SETS 

Sample maps with bandwidth of ~ 6.6 MHz (corre- 
sponding to the whole computational volume) of the in- 
tegrated overdensity, Thomson optical depth, and 21- 
cm differential brightness temperature ffuctuations are 
shown in Figure^ If the perturbations were small, linear 
and Gaussian, Stcs would be small. From the maps it is 
apparent that this is not the case. The large-bandwidth 
21-cm map is an excellent negative image of the optical 
depth map, indicating that 21-cm imaging can in fact be 
used for inferring the optical depth fluctuations. Com- 
paring equations ijSJ) and © we flnd 

<5re. = 0.0031(H-A^Hc) / ('5p(l+^)'/'-^^tfSr^)^- 

(9) 

The only differences between an actual t^s fluctuation 
map and a t^s fluctuation map derived from a 21-cm 
map will come from the first term above, i.e. from the 



3 






" 12.59C 



Fig. 1. — Maps of a single redshift output (z=12.9) showing integrated overdensity (left) integrated Thomson optical depth (middle) and 
integrated 21-cm emission (right) through the computational volume (6.6 MHz bandwidth). Each map is 47.5' on a side, with each pixel 
about 14". (Maps were produced using the package IFRIT of N. Gnedin). 



0.01 



n 



-0.01 



1 1 M^l^l 1 1 1 1 1 1 1 1 1 1 1 

- z=14.7 ! 


l^sjj . 1 1 1 1 1 1 1 1 1 1 1 1 1 

z=13.6 ~. 
1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 


1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 M 

^^^^^^^^^^ 

1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 


: X 

1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 



-10 10 



-10 10 
(5(ATj[mK] 



-10 10 



Fig. 2. — Contour plots from maps constructed from single redshift output showing integrated optical depth vs 21-cm integrated intensity. 
Shown is the distribution of map pixels, with contours showing a density of points that is 1, 10, and 100 points per cell. From left to 
right shows redshift 11.9, 12.6 and 13.6. The heavy line shows the expectation in the absence of fluctuations in the projected density map. 
Scatter comes from density fluctuations, while the change in slope (shown by the thin line) arises due to density-ionization correlations. 



density fluctuations along the the LOS, plus any nonlin- 
ear eff'ects. This gives a small signal, as Figure ^ shows, 
since there are many effective volumes along the LOS 
giving positive and negative contributions, which largely 
cancel each other. Otherwise, the fluctuation maps will 
be identical on a pixel- by-pixel basis. 

The ionization-density correlations are non-trivial. In 
Figure 12 we show a the distribution of 6tbs vs. 6{ATi,) 
for a few different redshifts. The fluctuations are strongly 
anti-correlated. Taking A^hc — 1-08 (i.e. assuming singly 
ionized helium everywhere hydrogen is ionized as done 
in the simulation), and assuming that the density fluc- 
tuations average to zero, equation (j^l gives the expected 
relation: S{ATb)[K] ^ 2.54 Stcs- The best-flt slope found 
in the simulations (shown with thin line in Figure [SJ 
differs from the simplest expectation (thick line) at a 
noticeable level. The scatter arises because the mean 
overdensity in the maps is not zero, as shown in Fig- 
ure Q Furthermore, these density fluctuations are corre- 
lated with optical depth and 21-cm emission. Physically 
this results from the inside-out, rather than random na- 
ture of the ionization process. The denser regions are 
on average ionized earlier, and are also more non-linear 
(as evidenced by the larger deviations in the upper left 
of each curve). The best fit slope varies with redshift, 
but within this ionization fraction range is never differ- 
ent from the theoretical value by more than 25%. This 



suggests that even with no understanding of the reion- 
ization process it will be possible to use 21-cm maps to 
reconstruct t^s to better than 25% accuracy. From the 
morphology of the 21-cm fluctuations, it should be possi- 
ble to understand the reionization process at a level that 
allows a much better understanding of this slope. 

The 21-cm fluctuations can thus be used to reconstruct 
the optical depth fluctuation map that leads to CMB 
polarization. With a CMB polarization map in hand of 
sufficient quality one could simply do a direct template 
search to determine the CMB quadrupole at a given lo- 
cation at the epoch of reionization. Residual contamina- 
tion will come from velocity-induced quadrupoles during 
reionization, which are expected to be roughly 10% of 
the primordial quadrupole. 

5. OBSERVATIONAL PROSPECTS 

The signals discussed here are small. The 21-cm emis- 
sion fluctuations on scales of order tens of arcseconds are 
of order a few mK. For wavelengths of several meters the 
diffraction limit corresponding to resolution of tens of 
arcseconds requires effective aperture sizes of tens of km. 
The LOS extent in frequency space is of order 1 MHz, 
and the svstem temperatur e is ~500K, set by the Galaxy 
fe.g. lFurlanetto et al.l20"0^ . For a single dish big enough 
to just resolve the sources (tens of km across), the noise 
scales as T^ys/V AvAt for bandwidth Ai/ and observing 



4 



time Ai. It would take roughly an hour to integrate down 
to 5 niK with this instrument. Using a smaller aperture 
dilutes the signal by the square of the telescope diame- 
ter, d, and thus increases the requisite integration time 
by d^. This will be a challenging signal that will require 
square kilometers of collecting area. 

Imaging the CMB polarization on scales of tens of arc- 
seconds to the requisite sensitivity will not be easy. The 
signal will be on the order of the primordial quadrupole 
times the Thomson optical depth, or roughly 0.1 fiK. 
The system temperatures at mm wavelengths are tens of 
Kelvin at best. A single telescope operating at mm wave- 
lengths with a 20 GHz bandwidth and a system temper- 
ature of 50K would require roughly six months of inte- 
gration time to reach 0.1 fiK sensitivity per diffraction- 
limited beam element. At 2mm, a 10" beam requires 
a 40m aperture. The Atacama Large Millimeter Array 
(ALMA) in a compact configuration roughly has this an- 
gular resolution and collecting area, with slightly more 
collecting area than is required, but an array filling factor 
that will resolve out some of the fiux. A rough estimate 
of ALMA observing times yields time scales of months 
of continuous integration. If giant ionized regions exist 
around large clusters of sources, then this could lead to 
signals that are an order of magnitude larger and all in- 
tegration times thus drop by two orders of magnitude. 
If such larger ionized regions are visible in 21-cm bright- 
ness fluctuations then ALMA will be able to image the 
CMB polarization directly. However, imaging the general 
morphology of reionization in 21-cm emission and CMB 
polarization will both be challenging undertakings. 

6. SUMMARY AND DISCUSSION 

We have shown that fluctuations in 21-cm emission 
from the epoch of reionization and CMB polarization 
fluctuations should trace each other extremely well. High 
quality redshifted 21-cm observations will be able to 
reconstruct the optical depth to Thomson scattering 
to better than 20% accuracy; better accuracy can be 
achieved with a better understanding of reionization. 

A map of the Thomson optical depth obtained from 
21-cm emission will allow at least two important im- 
provements in our knowledge. If reionization is a con- 
tributor to foreground "B mode" polarization anisotropy, 
this will allow cleaning of this foreground. Also, the po- 
larization signal can be used to measure the primordial 
quadrupole at the time of reionization. A coarse grid 
of all-sky coverage will allow reconstruction of the large 



scale structure in the universe; this includes the volume 
of the Dark Ages, which is currently unobservable due to 
the lack of sources. With dense coverage a reconstruction 
of nearly the entire primordial density field at a snapshot 
oi z ^ 1100 within our Hubble volume would be possi- 
ble; since each point only provides the quadrupole (a very 
large scale convolution) it is not clear that the reconstruc- 
tion would be accurate in the presence of astrophysical 
contaminants, noise, and bulk flows that contribute ve- 
locity quadrupoles. However, this signal contains unique 
information about the density fluctuations at z ^ 1100 
over a large fraction of our Hubble volume. 

Thomson scattering of the primordial quadrupole at 
lower redshifts has been discussed in the context of us- 
ing galaxy clusters as Thomson scatterers, but the 21-cm 
emission has several important advantages: 1) the Thom- 
son optical depth can be accurately reconstructed from 
the 21-cm map, while galaxy cluster optical depths may 
not be measurable at the requisite sensitivity; the optical 
depth derived from Sunyaev-Zel'dovich measurements is 
weighted by the Compt on t/-parameter an d is not appro- 
priate for this purpose ijKnox et al.ll2004() : 2) the overlap 
in information of the CMB quadrupole at z 10 and 
our measured CMB sky is much reduced when compared 
with clusters at 2; ~ 0.2 — 0.5; local clusters are better as 
probes of the ISW effect; 3) the Sunyaev-Zel'dovich effect 
and gravitational Icnsing of the CMB will not be corre- 
lated with the CMB polarization signal and one does not 
expect radio halos, quasar and galaxy overdensities, or 
strongly lensed background objects. 

With only a polarization map one could still measure 
the direction of the CMB quadrupole, which would al- 
low tests of isotropy and homogeneity of the CMB at 
z ^ 10, provided the reionization process is relatively 
sharp in redshift. Large ionized regions around quasars 
would provide interesting targets for ALMA to try to 
measure this signal, although imaging a field to fiK sen- 
sitivity with a quasar at the center will require some 
care. While the signals will be difficult to image with 
high signal to noise, these observations offer tremendous 
opportunities for gaining information about our universe 
that is not available by any other means. The hurdles are 
not insurmountable and the observations required (high 
signal to noise neutral hydrogen maps and high resolu- 
tion imaging of the CMB polarization fluctuations) are 
already on the ultimate wish list for physical cosmology. 

GPH is supported by an NSERC Discovery grant. We 
thank Olivier Dore and Ue-Li Pen for useful discussions. 



REFERENCES 



Baumann, D. & Cooray, A. 2003, New Astronomy Review, 47, 839 
Bunn, E. F. 2006, Phys. Rev. D, 73, 123517 

Furlanetto, S., Oh, S. P., & Briggs, F. 2006, ArXiv Astrophysics 

e-prints 
Hu, W. 2000, ApJ, 529, 12 
Iliev, I. T., et ah 2006a, MNRAS, 873 

Iliev, I. T., Mellema, C, Pen, U.-L., Merz, H., Shapiro, P. R., & 

Alvarez, M. A. 2006b, MNRAS, 369, 1625 
Iliev, I. T., Pen, U.-L., Bond, J. R., Mellema, C, & Shapiro, P. R. 

2006c, New Astronomy Reviews, in press I astro-ph/0607209 I 
Iliev, I. T., Shapiro, P. R., Ferrara, A., & Martel, H. 2002, ApJL, 

572, L123 

Kamionkowski, M. & Loeb, A. 1997, Phys. Rev. D, 56, 4511 
Knox, L., Holder, G. P., & Church, S. E. 2004, ApJ, 612, 96 
Mason, B. S. & Myers, S. T. 2000, ApJ, 540, 614 



Mellema, G., Iliev, I. T., Alvarez, M. A., & Shapiro, P. R. 2006a, 

New Astronomy, 11, 374 
Mellema, C, Iliev, I. T., Pen, U. L., & Shapiro, P. R. 2006b, 

MNRAS, in press ( astro-ph/0603518 I 
Merz, H., Pen, U.-L., & i'rac, H. 2005, New Astronomy, 10, 393 
Mortonson, M. J. & Hu, W. 2006, ArXiv Astrophysics e-prints 
Portsmouth, J. 2004, Phys. Rev. D, 70, 063504 
Sazonov, S. Y. & Sunyaev, R. A. 1999, MNRAS, 310, 765 
Seto, N. & Pierpaoli, E. 2005, Physical Review Letters, 95, 101302 
Seto, N. & Sasaki, M. 2000, Phys. Rev. D, 62, 123004 
Shimon, M., Rephaeli, Y., O'Shea, B. W., & Norman, M. L. 2006, 

MNRAS, 368, 511 
Spergel, D. N. e. a. 2003, ApJS, 148, 175